Don Cox wrote:

Clicks are not "high frequency". They are impulse signals and as such
contain all the frequencies the system can reproduce in roughly equal
amounts, as does hiss.

You cannot remove clicks by processing in the frequency domain. They
have to be detected individually and the signal then edited or patched,
which is only easy to do with digital methods.

Bradleys reply:

Yes, that sounds right. Will Don Cox or another can assist us in
understanding something here. Is the idea to get the 78rpm record recording
into a digital signal domain to achieve a restoration of maximum quality? We
thought that converting to digital and signal processing the retrieved
signal would be a preferred path. Is rather the idea to maintain analog
playback and to be able to play these old records on a turntable although
with perhaps a modern stylus and arm?

Would this not even work in the analog domain, perhaps in real time while
playing the record through (ultimately) a mono-channel? The idea similarly
applies to mono 33 1/3 rpm records but fails when one moves to stereo age

It seemed plausible that once an inverted signal of the snap and pop was
combined with the program signal that the resulting error might consist of
mainly higher frequency noise, that lower frequency components would suffer
less from phase distortions and other variances between the retrieved
channels (refer back to the original post on this). Also, the pop and crack
should be able to fairly narrowly define a time where the problem was. By
perhaps applying a correction only during that gated window less attenuation
and disturbance of the program would result.

We erred in speaking of the issue being only a high frequency problem
though. Impulse signals are spectrally very broad in frequency distribution.
Still there is a lot of high frequency energy in a pop. Hearing impairment
keeps us from being able to clearly discern consonant sounds in speech
resulting in such sounds as B and V sounding the same. It is the highest
audible frequencies that define the differences in how these sounds sound
and so discrimination of consonant sounds is usually the first thing to be
lost in human hearing. Seeing the lips of a person speaking--the way the
lips move when these sounds are made is very different enables a decently
intelligible hearing. We first know we have a problem when trying to
communicate over the phone--no lips.

The reason we suggested that the noise signal may be somewhat duplicated in
the non-program channel is that the response of a stylus and cartridge to a
crack or scratch in a record might be similar in both channels and certainly
the time of the impulse should be nearly identical. If this actually is the
case in practice I would expect that someone has known and that this might
be a common way to improve the extracted program sound. If such procedures
are not known it may be because it doesn't work. Just in case no one has
thought of trying it maybe someone would and let us know about it. If stylus
and cartridges exist that detect side variation in one channel and vertical
variation in the other that would seem to be the right type for the
experiment. Knowing nearly nothing about sound recorded in record grooves,
this could be entirely nonsense. A broader comment/response on these ideas
would help us out here.